JP2000513129A - Method of tracking incomplete writes in a disk array, and disk storage system performing such a method - Google Patents

Abstract

SUMMARY OF THE INVENTION A method for tracking incomplete writes in a disk array uses a plurality of write commands (t1, t2, t3, t4, t5) to identify each block in the array to be written. Write), and generating a list (16) of extended write areas (eg, PB8, 289 of the list at time t1) for only the most recently received write command; Each extended write area includes the block to be written and any additional blocks that may be written by a later write command (eg, PB8,289 + 8 includes writing PB8,290 at t1). ), Each time a write command to write a particular block that is not in any of the extended write areas in the list (eg, write at t4) is received, one extension Modifying the list by replacing the area (eg, the last entry in the list at t3) with a new extended write area containing the particular block (eg, the first entry in the list at t4); And storing a copy of the list (17 of t4) on a magnetic medium each time the step of performing is performed.

Description

Description: Title: A method for tracking incomplete writes in a disk array and a disk storage system performing such a method Background of the Invention: The present invention relates to a method for tracking incomplete writes in a disk array and a disk storage system performing such a method. In the prior art, the term "RAID" disk array has been defined to mean a redundant array of inexpensive disks, and several different RAID disk arrays have been defined. These include a level 1 RAID disk array, a level 3 RAID disk array, and a level 5 RAID disk array. "A Case for Redundant Arrays of Inexpensive Disks (RAID)", by Patterson. Et al., Report No. UCB / CSD 87/391, December 1987. See the Computer Science Division of the University of California at Berkeley, Mon, Berkeley. In a level 5 RAID disk array, both parity and data are striped across a set of several disks. FIG. 1 shows an example of a level 5 RAID disk array in which an array is formed by a set of five disks labeled as disk 0, disk 1,. Each column of the array contains data and parity stored in a single disk in the set. Each row of the array contains data and parity striped across all five disks in the set. In FIG. 1, each row of the array consists of one parity chunk on one disk and four data chunks on the other four disks. Also, each data chunk and each parity chunk are divided into several physical blocks. A single block is the smallest part of a chunk that can be separately addressed by a user program with a read or write command. In FIG. 1, there are eight blocks per chunk. Each block consists of a predetermined number of bytes (eg, 512 bytes) and one cyclic redundancy check byte called a "CRC" byte. In the array of FIG. 1, block 0 of row 0 is addressed by a logical address 0 read / write command. As the logical addresses are sequentially incremented by one, the data blocks are block 1-7 of data chunk 0, block 0-7 of data chunk 1, block 0-7 of data chunk 2, and block 0 of data chunk 3. -7, blocks 8-15 of data chunk 4, blocks 8-15 of data chunk 5, etc. in that order. For example, block 8 of data chunk 5 has logical address 40. When writing to a data block, the CRC bytes within that block are also generated and written. In addition, a parity block having the same block number as the data block is also generated and written. This parity block is written using odd or even parity. In the even parity, the exclusive OR of the parity block and all the data blocks having the same block number is a block of which all are 0. Conversely, in the odd parity, the exclusive OR of the parity block and all the blocks having the same block number becomes one block. One way to generate a new parity block for a new data block to be written is as follows. First, the existing data blocks and their parity blocks are read from their respective disks. Next, an exclusive OR of the read parity block and the read data block is calculated, and a new parity is calculated by calculating an exclusive OR of the logical block and the new data block. This new parity block and the new data block are then written to their respective disks. During execution of the read command, the CRC byte is regenerated from the block of data to be read. If the regenerated CRC byte is different from the stored CRC byte, the block of data to be read contains an error. To correct this error, the erroneous data block is read by: a) reading all the other blocks (data and parity) on the disk having the same block number as the erroneous data block; b) exclusion of those blocks It is regenerated by taking the logical OR. Here, a case is considered in which the execution of a specific write command that attempts to write data to a block having a specific block number “i” is started, but the execution is interrupted before completion. Such an interruption may occur, for example, due to a power outage. In the above case, the interruption may occur after the writing of the new data block is completed but before the writing of the new parity block begins. Similarly, an interruption can occur after the writing of a new parity block is completed but before the writing of a new data block begins. In either case, the exclusive OR of all blocks with block number "i" is not equal to all zeros or all ones. At the same time, the ECC bytes for all blocks with block number "i" will be correct. After the cause of the interruption has been resolved, the array will continue to be read and written by the user program. If any data block with block number "i" is read and the ECC byte detects an error, the remaining data block with the same block number and the parity block are XOR'd with the error to find an error. An attempt is made to regenerate the data block. However, the regeneration process does not work because of the previous incomplete writing. This problem has been addressed in the prior art by providing a flag in the disk array that is set to "1" when the array starts operating and reset to "0" when the array stops operating in the normal manner. . Thus, if the flag is set to "1" before the flag is set when the array begins to operate, then normal operation of the array must have been previously suspended. However, a disadvantage of the above prior art flags is that it takes too long to identify a particular block that has been incompletely written after the flag is found to be in the wrong state. To find the incomplete block of the write, all data blocks and all parity blocks of the entire array must be read, and therefore the parity blocks must be recalculated from the read data blocks, and The parity block read must be compared with the parity block read. For large arrays, this process can take an entire day to complete. Accordingly, it is a primary object of the present invention to provide a method of tracking incomplete writes in a disk array and to perform such a method to check the parity for all blocks of the entire array after normal operation of the array has been interrupted. To provide a disk storage system that eliminates the need to do so. BRIEF SUMMARY OF THE INVENTION: In the present invention, incomplete writes in the disk array are tracked by a control process in the digital computer. The control program sequentially receives a plurality of write commands identifying each block in the array to be written. In response, the control program generates an expanded list of write areas for only the most recently received write command. Each such extended write area includes a block to be written and additional associated blocks that may be written by a subsequent write command. The list is changed by the control program each time a write command for writing a specific block not in the extended write area in the list is received later. In this change, one extended write area in the disk is replaced by a new extended write area containing that particular block. In addition, each time a change step occurs, the control program stores a copy of the list on magnetic media. Due to an unexpected interruption, such as a power outage, the write operation required by any of the received write commands may be initiated but cannot be completed. In the present invention, the presence of an incomplete write is detected by checking only the parity of the block identified in the copy of the list. As a result, a considerable amount of time is saved relative to the time required to check the parity of every block in the entire disk array. As a practical numerical example of the above savings, consider the case where the entire array stores 900 gigabytes and the extended write area in the replica amounts to a total of 500 megabytes. It takes about 10 minutes to check the parity of the 500 megabyte portion of the disk array. However, 900 gigabytes is 1800 times larger than 500 megabytes, and it would take about (1800) * (10 minutes) or about 30 hours to check the parity of the entire array. Increasing the total number of extended write areas in the list increases the likelihood that the list will contain all the blocks to be written by the write command. This is a desirable result because it reduces the number of times that a copy of the list needs to be updated on magnetic media. Preferably, less than 1% of write commands require addition to the list of new extended write areas. BRIEF DESCRIPTION OF THE DRAWINGS: FIG. 1 shows the structure of a level 5 RAID disk array in which incomplete writing can occur. FIG. 2 illustrates one preferred embodiment of a disk storage system that tracks incomplete writes in the array of FIG. 1 in accordance with the present invention. FIG. 3 shows a preferred structure of the extended write area used by the disk storage system of FIG. 2 to track incomplete writes. FIG. 4 illustrates how the extended write area of FIG. 3 is stored in a list and a copy of the list in the disk storage system of FIG. FIG. 5 illustrates a second preferred structure of the extended write area used by the disk storage system of FIG. 2 to track incomplete writes. Detailed description: One preferred embodiment of a disk storage system for tracking incomplete writes in a disk array according to the present invention is shown in FIG. The system of FIG. 2 includes a set of five disks (disk 0 through disk 4), which store data blocks and parity blocks in a level 5 RAID disk array as described in the background with respect to FIG. The system of FIG. 2 also includes a digital computer 11 coupled to the disk array, a semiconductor memory 12 coupled to the computer 11, and a console 13 coupled to the computer. Stored in the memory 12 are a control program 14 and a plurality of user programs, one of which is shown as a user program 15j. Each user program includes a read / write command that includes a logical address that identifies a particular data block in the disk array to be read or written. These read / write commands are sequentially sent to a queue 14a in the control program 14 for execution. Each write command in queue 14a is later received by control program 14 and is performed as follows. First, the control program analyzes the logical address in the write command to determine the number of each data block to be written, the number of the disk containing the data block, and the number of the disk containing the corresponding parity block. I do. Next, the control program reads the current contents of the aforementioned data blocks and their corresponding parity blocks from their respective disks. Next, the control program calculates an exclusive OR of the read parity block and the read data block, and calculates a new parity block by taking an exclusive OR of the read parity block and the new data block. Next, the control program writes a new parity block and a new data block to their respective disks. Thus, the execution of each write command requires that two read operations be performed on two different disks, after which the two write operations are performed on the same two disks. These two write operations may be performed in any order, so that if an unexpected interruption, such as a power outage, occurs, the write operation on the disk will be performed before the write operation on the other disk begins. Cannot be completed. To track such imperfect write operations in the disk array, the disk storage system of FIG. 2 further includes a list 16 in semiconductor memory 12. This list 16 identifies extended write areas for those write commands most recently received by the control program 14 from the queue 14a. Each extended write area in the list contains the block to be written by the write command and additional related blocks. FIG. 3 described later shows two specific examples of these extended writing areas. Each time a write command is received from queue 14a for execution, control program 14 compares the identity of the block to be written to the identity of the block in the extended write area in list 16. If the block to be written is not in any of the extended write areas in list 16, control program 14 places one extended write area in list 16 into a new extended write area containing the particular block to be written. Replace with the area. Preferably, all of the extended write areas in the list 16 are arranged in the list in such a manner that the oldest used ones are deleted. In this case, when a new extended write area needs to be added to the list, the one extended write area that is replaced is the oldest used. In any case, the extended write area is not removed from list 16 until all received write commands that write blocks in the extended write area have been completely executed. In the disk storage system of FIG. 2 as well, a copy of list 17 is stored on magnetic media. In one embodiment, replica 17 is stored on one of five disks holding a level 5 RAID disk array. Alternatively, the disk storage system of FIG. 2 may include an auxiliary magnetic medium 18 on which a copy 17 of the list is stored. This replica 17 differs from the list 16 in that the extended write areas it identifies are not arranged in a particular order. However, both the list 16 and the replica 17 identify the same extended write area. One particular structure for the extended write area stored in the list 16 and its replica 17 is shown in FIG. Here, the large arrow 21 represents a write command received from the queue 14a by the control program 14 requesting writing of the two data blocks on which the arrows are drawn. These data blocks are stored on disk 2 in physical blocks PBY + 5 and PBY + 6. The extended write area for those two blocks is indicated by hatch line 22 in FIG. 3, which includes physical blocks PBY + 5, PBY + 6, PBY + 7, PBY and PBY + 1 on all disks. Similarly, in FIG. 3, another large arrow 25 is drawn in one data block to be written in response to the second write command from queue 14a. The data block is stored on the disk 4 as a physical block PBY + 6. The extended write area for that block is indicated by dot 26 in FIG. 3, which includes physical blocks PBY + 6, PBY + 7, PBY + 8 and PBY + 9 on the entire disk. In FIG. 3, the range of the extended writing area indicated by hatch line 22 is determined as follows. First, the control program generates three extended logical addresses by adding 1, 2, and 3 to the logical address of the block to which the write command 21 last writes. Next, the control program identifies physical blocks corresponding to those extended logical addresses. Reference numbers 23a, 23b and 23c indicate the results of this step. In this case, the extended write area is defined to consist of those physical blocks on all disks and the blocks to be written. The extended writing area indicated by the dot 26 is determined in the same manner. In FIG. 3, reference numerals 27a, 27b, and 27c indicate the results of a) adding 1, 2, and 3 to the logical address of the block to which the write command 25 writes, and b) identifying the corresponding physical block. In the list 16, the extended writing area indicated by the hatch line 22 is identified in a simple form such as (PBY + 5, +2) and (PBY, +1). This compact form reduces the size of the list. In this form, the first item in parentheses is a physical block on all disks in the extended write area, and the second item is a count of the number of physical blocks following the block on all disks. Similarly, the extended writing area indicated by the dot 26 is represented in a simple form such as (PBY + 6, +3). Here, it is assumed that the write command 25 is received from the queue 14a while the extended write area indicated by the hatch line 22 is in the list 16. In that case, the extended writing area indicated by the dot 26 is not added to the list 16. This is because the control program 14 compares the identity of the block to be written with the identity of the block in the extended write area of the list 16 as described above. If all of the blocks to which a particular write command writes are in the extended write area of list 16, no new extended write area for that command is added to the list. FIG. 4 shows a specific example of the operation of the list 16 and its copy 17. In this example, the initial contents of list 16 and its copy 17 occur at time t0. At that time, the most recently used extended write area in the list 16 is (PB35,191 + 6), the next most recently used extended write area is (PB765 + 6), and the oldest recently used extended write area is The embedding area is (PB136, 353 + 6). Thereafter, at time t1 in FIG. 4, a write command for writing the physical blocks PB8, 290 is received by the control program 14 from the queue 14a. This physical block is included in the extended write area (PB8, 289 + 8) in the list 16. Therefore, the new extended write area is not added to the list, so the copy 17 of the list remains unchanged. However, the extended write areas in list 16 are rearranged to maintain their order of use. Next, at time t2 in FIG. 4, a write command for writing the physical blocks PB136 and 354 is received by the control block 14 from the queue 14a. This physical block is included in the extended write area (PB136, 353 + 10) in the list 16. Therefore, the new extended write area is not added to the list, and the copy 17 of the list remains unchanged. However, the extended write area in the list 16 is again rearranged to maintain its use order. Next, at time t3 in FIG. 4, a write command for writing the physical block PB767 is received by the control program 14 from the queue 14a. This physical block is included in the extended write area (PB765 + 6) in the list 16. Therefore, the new extended write area is not added to the list and the copy 17 of the list is not changed here either. However, the extended write areas in list 16 are rearranged to maintain their order of use. Next, at time t4 in FIG. 4, a write command for writing the physical blocks PB91 and 532 is received by the control program 14 from the queue 14a. This physical block is not included in any of the extended write areas in the list 16. Therefore, a new extended write area (PB91, 532 + 3) is added to the list 16, the oldest used extended write area (PB524, 911 + 6) is deleted from the list 16, and all extended write areas in the list are deleted. They are arranged in order of use. Also, the copy 17 of the list is written so that its contents are the same as the new list 16. Finally, at time t5 in FIG. 4, a write command for writing the physical blocks PB136, 356 is received by the control program 14 from the queue 14a. This physical block is included in the extended write area (PB136, 350 + 10) in the list 16. Therefore, the new extended write area is not added to the list and the copy of the list 17 remains unchanged. However, the extended write area in the list 16 is again rearranged to maintain its use order. Note that for simplicity of FIG. 4, a total of five extended write area keys 16 (and their duplicates 17) are not shown. However, the list is much larger in a real disk storage system. Preferably, the extended write area in the disk 16 covers the entire storage area in the 100 megabyte to 1 gigabyte array. Increasing the total number of extended write areas in the list 16 increases the likelihood that the list contains all blocks to which the write command from the queue 14a writes. This is a desirable result because it reduces the number of times that the list replicas 17 need to be updated on magnetic media. Preferably, less than 1% of write commands from queue 14a require the addition of a new extended write area to list 16. However, increasing the total number of extended write areas in list 16 also increases the number of blocks whose parity needs to be checked after an unexpected interruption. For example, assume that an unexpected interruption occurs while the system of FIG. 2 is executing a write command from queue 14a. Because of the interruption, writing a data block to one disk may be completed before writing the corresponding parity block to another disk, or vice versa. As a result, the array must check for the presence of such incomplete writes after the cause of the interruption has been resolved. In the present invention, the presence of an incomplete write is detected by checking only the parity on the block identified in the replica 17 of the list 16. As a result, a significant amount of time is saved relative to the time required to check the parity on all blocks of the entire level 5 RAID disk array. As a practical numerical example of the above savings, consider the case where the entire array stores 900 gigabytes and the extended write area in replica 17 totals 500 megabytes. It takes about 10 minutes to check parity for a 500 megabyte portion of the disk array. However, 900 gigabytes is 1800 times larger than 500 megabytes, and checking parity for the entire array takes about (1800) * (10 minutes), or about 30 hours. One preferred method of tracking incomplete writes in a disk array and a preferred disk storage system for performing such a method have been described in detail herein. However, various changes may be made in those details without departing from the scope of the invention. For example, in FIG. 3, the extent of each extended write area was determined by adding 1, 2, and 3 to the logical address of the block to which the write command from queue 14a last writes. However, as a modification, the number of integers added to the logical address may be increased. By increasing in this way, the possibility that the extended write area includes all blocks to be written by a later write command from the queue is increased. Preferably, the number added to the logical address of the last write block ranges from 1 to 200. As another variation, the extent of each extended write area may be determined by subtracting 1, 2, ... from the logical address of the block to which the write command from queue 14a first writes. In this variation, the extended write area includes the blocks in the array that precede the block to be written. This also increases the likelihood that the extended write area will contain all blocks to which future write commands from the queue will write. Preferably, the number subtracted from the logical address of the first block to be written is in the range of 1 to 200. As another variation, each extended write area in list 16 may be configured to add 1, 2,... To the logical address of only the block to which the write command is writing and the last block to be written, and / or Or additional blocks identified by subtracting 1, 2,... From the logical address of the first block to be written. An example of this change is shown in FIG. In FIG. 5, an arrow 35 indicates a logical address LA. _w And LA _w Represents a write command from queue 14a seeking to write two blocks with +1. These two blocks exist on the disk 4 as physical blocks PB Y + 5 and PB Y + 6. For those two blocks, the extended write area is identified in FIG. 5 by adding 1, 2,... 15 to the logical address of the block to be written and the last block to be written. Additional blocks are defined as being additional blocks identified by subtracting 1, 2,... 8 from the logical address of the first block to be written. These additional blocks are designated in FIG. 5 by reference numbers 35 (+1) to 35 (+15) and 35 (-1) to 35 (-8). In the list 16 and its copy 17, the extended write area in FIG. _w -8, LA _w +16). In this form, the first item in parentheses is the logical address of the first block in the extended write area, and the second item is the logical address of the last block in the extended write area. If the extended write area is defined as shown in FIG. 5, after an unexpected interruption, the presence of an incomplete write is detected as follows. First, the control program 14 converts all logical addresses included in the write extension area in the copy 17 into corresponding physical blocks. During this conversion, the particular disk on which each physical block resides is ignored. Next, those physical blocks are read from all disks in the array. Parity is regenerated from the read data block and compared with the read parity block. As another modification, the size of the extended writing area in FIG. 5 may be made smaller or larger. Preferably, the number added to the logical address of the last block to be written and the number subtracted from the logical address of the first block to be written are in the range of 1 to 200. Further, as another modification, the present invention is not limited to the arrangement in which the extended writing areas are arranged so that the list 16 is deleted in the oldest used order. Alternatively, any predetermined algorithm may be used to remove one extended write area from list 16 to make room for a new extended write area. As an example, the extended writing area removed from the list 16 may be the longest in the list. As another example, the extended writing area to be removed from the list 16 may be randomly selected. As a further modification, the list 16 need not be stored in the same semiconductor memory 12 holding the control program 14 as shown in FIG. Alternatively, the list 16 may be stored in a memory included in the computer 11 or in a set of registers. As yet another modification, the RAID disk array holding the data blocks and the parity blocks may not be a level 5 array. Alternatively, the data blocks and parity blocks may be arranged on the disk in any way that allows the erroneous data blocks to be regenerated from the parity blocks. For example, the array holding the data blocks and parity blocks may be a level 3 RAID disk array as described in the "Background" section. Therefore, it is to be understood that the scope of the invention is not limited to the details of any of the specific embodiments described, but is defined by the appended claims.

[Procedure amendment] [Date of submission] December 1, 1999 (1999.12.1) [Contents of amendment] (1) The following is placed between page 3, line 29 and page 4, line 1 Insert The problem addressed by WO-A-94 / 29795 is the same as in the present invention. However, the solution of WO-A-94 / 29795 is substantially different from the present invention. That is, while the present invention envisages creating an extended write area containing each block in normal disk storage space to form part of an integrated, striped data / parity structure. This is a solution of WO-A-94 / 29795 based on providing an additional NV RAM area separate from the usual disk array structure to store a list of possibly inconsistent blocks. Different from the law. The proposed solution has the advantage of reducing the size of the list and effectively identifying all blocks associated with inconsistencies in the disk array. (2) Insert “The claimed invention is described below.” Between the fifth and sixth lines of page 4 of the specification. (3) The claims are corrected as shown in the attached sheet. Claims 1. A method performed by a computer program to track incomplete writes in a disk array (see FIG. 4) , comprising a plurality of write commands ( FIG. 4) identifying each block in said array to be written. 4) sequentially receiving write commands at t1, t2, t3, t4, and t5 in FIG. 4; and an extended write area ( in the list at time t1) for only the most recently received write commands . Generating a list ( e.g., PB8, 289) (16 in FIG. 4) , wherein each extended write area includes a block to be written and additional blocks associated therewith (e.g., PB8, 299 + 8). the PB8, and a write at 290 t1), any extended writing area (e.g. a particular block nor in writing) at t4 in said list Each time a write command for performing write is received later, one extended writing area (e.g., the last entry in the list 16 of t3) New Additional write area including the specified block (e.g., at t4 by replacing the first entry) in the list and a step of storing and changing the list, replication of said list on a magnetic media each time step of the change is performed (17 t4), Method. 2. The method of claim 1, wherein the extended area in the list covers 100 megabytes to 1 gigabyte of the entire storage area in the array. 3. The method of claim 1, wherein each extended writing area includes between one and 500 additional blocks. 4. The method of claim 1, wherein each extended write area includes the same physical block on all of the disks. 5. The method of claim 1, wherein each extended write area includes a selected physical block on a selected disk. 6. The method of claim 1, wherein the altering step replaces an oldest written extended write area in the list. 7. The method according to claim 1, wherein the program generates the list in a volatile semiconductor memory. 8. The method of claim 1, wherein the program generates the list in a set of volatile registers. 9. The method of claim 1, wherein the magnetic medium on which the copy of the list is stored is a disk in the array. 10. The method of claim 1, wherein the magnetic medium on which the copy of the list is stored is an auxiliary medium external to the disk array. 11. The method of claim 1, wherein the array is a level 5 RAID disk array. 12. a) starting to write a particular block in the array, identified by the received write command; b) suspending the writing of the particular block before it has been completely written; The method of claim 1, further comprising: 13. a) reading the replica from the magnetic medium after writing of a particular block in the array has been interrupted; and b) checking parity only on blocks included in the extended write area in the replica. The method of claim 1, further comprising: 14． An array of disks coupled to a digital computer, and a program in the computer for sequentially receiving a plurality of write commands identifying respective blocks in the array to be written, and a plurality of most recently received writes. A list in the computer of an extended write area for commands only, wherein each extended write area includes a block to be written and additional blocks associated therewith, wherein the program comprises: By replacing a single extended write area with a new extended write area containing the specific block each time a write command to write a particular block that is not in any extended write area is subsequently received. Modifying the list, wherein the list is modified each time the list is modified by the program. A disk storage system further comprising a magnetic medium for storing a copy of the disk.

Claims (1)

[Claims] 1. A computer program that tracks incomplete writes in a disk array. Is a method performed by   A plurality identifying each block in the array to which writing is to be performed Sequentially receiving write commands of   List of extended write areas for only the most recently received write commands Generating each of the extended write areas in a block in which writing is to be performed. And additional blocks associated with it,   Write a specific block not in any extended write area in the list Each time a subsequent write command is received, one extended write area is Modify the list by replacing it with a new extended write area containing a lock Steps and   Storing a copy of the list on a magnetic medium each time the changing step is performed And the step of causing. 2. The extension area in the list is the array from 100 megabytes to 1 gigabyte. 2. The method of claim 1, wherein the entire storage area is covered. 3. Each extended write area includes between 1 and 500 additional blocks Item 1. The method according to Item 1. 4. Each extended write area contains the same physical block on all of the disks, The method of claim 1. 5. Each extended write area contains the selected physical block on the selected disk, The method of claim 1. 6. The step of modifying includes the oldest written extension in the list. The method of claim 1, wherein the writing area is replaced. 7. 2. The program according to claim 1, wherein the program generates the list in a volatile semiconductor memory. The method described in. 8. The program generates the list in a set of volatile registers. 2. The method according to 1. 9. The magnetic medium on which the copy of the list is stored is a disk in the array The method of claim 1, wherein 10. The magnetic medium on which the copy of the list is stored is a copy of the disk array. The method of claim 1, wherein the method is an external auxiliary medium. 11. The array of claim 1, wherein the array is a level 5 RAID disk array. Method. 12. a) a particular command in the array identified by the received write command Initiating the writing of a block, b) writing the particular block Interrupting before being completely written. Law. 13. a) the magnetic medium after writing of a particular block in the array is interrupted; And b) reading the copy from the extended write area in the copy. Checking parity on lock only. Method. 14． An array of disks coupled to a digital computer Multiple write commands to identify each block in the array to be sequenced. The next program to be received in the computer and the most recently received programs A list in the computer of an extended writing area only for embedded commands. Each extended write area is a block to be written and additional blocks associated with it The program is not in any extended write area in the list Each time a write command to write a particular block is received later, By replacing the extended write area with a new extended write area including the specific block. The list is changed by the program, and the list is changed by the program. A disk drive, further comprising a magnetic medium for storing a copy of the list each time it is changed. Memory system.